Edit Page

Working with queries

Queries (aka query instances) are the chainable deferred objects returned from model methods like .find() and .create(). They represent a not-quite-yet-fulfilled intent to fetch or modify records from the database.

var query = Zookeeper.find();

The purpose of query instances is to provide a convenient, chainable syntax for working with your models. Methods like .populate(), .where(), and .sort() allow you to refine database calls before they're sent down the wire. Then, when you're ready to fire the query off to the database, you can just await it.

If you are using an older version of Node.js that does not support JavaScript's await keyword, you can use .exec() or .then()+.catch(). See the section on "Promises and Callbacks" below for more information.

Most of the time, you won't think about query instances as objects per se, but as just another part of the syntax for communicating with the database. In fact, you may already be using these objects in your Sails app! If so, the following syntax should look familiar:

var zookeepers = await Zookeeper.find();

In this example, the call to Zookeeper.find() returns a query instance, but doesn't actually do anything until it is executed using the await keyword, and then the result is assigned to the zookeepers variable.

How it works

#

When you execute a query using await, a lot happens.

await query;

First, the query is "shaken out" by Waterline core into a normalized query. Then it passes through the relevant Waterline adapter(s) for translation to the raw query syntax of your database(s) (e.g. Redis or Mongo commands, various SQL dialects, etc.) Next, each involved adapter uses its native Node.js database driver to send the query out over the network to the corresponding physical database.

When the adapter receives a response, it is marshalled to the Waterline interface spec and passed back up to Waterline core, where it is integrated with any other raw adapter responses into a coherent result set. At that point, it undergoes one last normalization before being passed back to "userland" (i.e. your code) for consumption by your app.

Error handling

#

You can use a try/catch to handle specific errors, if desired:

var zookeepersAtThisZoo;
try {
  zookeepersAtThisZoo = await Zookeeper.find({
    zoo: req.param('zoo')
  }).limit(30);
} catch (err) {
  switch (err.name) {
    case 'UsageError': return res.badRequest(err);
    default: throw err;
  }
}

return res.json(zookeepersAtThisZoo);

The specific kinds of errors you could receive vary based on what kind of query you are executing. See the reference docs for the various query methods for more specific information.

Promises and callbacks

#

As an alternative to await, Sails and Waterline provide support for callbacks and promise-chaining. In general, you should use await whenever possible; it leads to simpler, easier-to-understand code, and helps prevent DDoS vulnerabilities and stability issues that can arise from throwing uncaught exceptions in asynchronous callbacks. That said, sometimes it is necessary to maintain backwards compatibility with an older version of Node.js. For this reason, all queries in Sails and Waterline expose an .exec() method.

Zookeeper.find().exec(function afterFind(err, zookeepers) {

  // Careful!  Do not throw an error in here without a `try` block!
  // (Even a simple typo or null pointer exception could crash the process!)

  if (err) {
    // uh oh
    // (handle error; e.g. `return res.serverError(err)`)
    return;
  }

  // would you look at all those zookeepers?
  // (now let's do the next thing;
  //  e.g. `_.reduce(zookeepers, ...)` and/or `return res.json(zookeepers)`)
  // …
});
//
// (don't put code out here)

As shown in the example above, the query is not executed right away, but notice that instead of using await to execute the query and wait for its result, we use the traditional .exec() method with a callback function. With this usage, we cannot rely on try/catch and normal error handling in JavaScript to take care of our errors! Instead, we have to manually handle them in our callback to .exec(). This style of error handling is the traditional approach used in Node.js apps prior to ~Summer 2017.

Under the covers, Sails and Waterline also provide a minimalist integration with the Bluebird promise library, exposing .then() and .catch() methods.

Zookeeper.find()
.then(function (zookeepers) {...})
.catch(function (err) {...});
//
// (don't put code out here)

In this example, the callback passed into .catch() is equivalent to the contents of the if(err) {} block from the .exec() example above (e.g. res.serverError()). Similarly, the .then() callback is equivalent to the code below the if(err) {} and early return.

If you are a fan of promises and have a reasonable amount of experience with them, you should have no problem working with this interface. However if you are not very familiar with promises, or don't care one way or another, you will probably have an easier time working with .exec(), since it uses standard Node.js callback conventions.

If you decide to use traditional promise chaining for a particular query in your app, please make sure that you provide callbacks for both .then() and .catch(). Otherwise errors could go unhandled, and unpleasant race conditions and memory leaks could ensue. This is not just a Sails or Waterline concept. Rather, it's something to be aware of whenever you implement this type of usage in JavaScript—particularly in Node.js—since unhandled errors in server-side code tend to be more problematic than their client-side counterparts. Omitting .catch() is equivalent to ignoring the err argument in a conventional Node callback, and it is similarly insidious. In fact, this is hands-down one of the most common sources of bugs for Node.js developers of all skill levels.

Proper error handling is particularly easy to neglect if you're new to asynchronous code. Once you've been at it for a while, you'll get in the habit of handling your asynchronous errors right after (or even better, right before) you write code that handles the successful case. Habits like this immunize your apps to those common bugs discussed above.

(Better yet: just use await!)

Notes

#
  • A query instance is not exactly the same thing as a Promise, but it's close enough for our purposes. The difference is that a query instance in Sails and Waterline is actually a Deferred, as implemented by the parley library. That means it doesn't start executing immediately. Instead, it only begins executing when you kick it off with either await, .exec(), .then(), or .toPromise().
  • A Node-style callback can be passed directly as a final argument to model methods (e.g. .find()). In this case, the query will be executed immediately, and model methods will not return a query instance (instead, the Node-style callback you provided will be triggered when the query is complete). Unless you are doing something very advanced, you are generally better off sticking to standard usage; i.e. calling .exec() or calling .then() and .catch().

Is something missing?

If you notice something we've missed or could be improved on, please follow this link and submit a pull request to the sails repo. Once we merge it, the changes will be reflected on the website the next time it is deployed.

Reference

Reference